Research on the Competitive Consequences of Common Ownership: A Methodological Critique

Abstract

This article argues that the evidence presented in several critiques of Azar, Schmalz, and Tecu’s (AST) “airlines” paper does often not back the conclusion these studies draw. Specifically, widely circulated studies claiming that there are no anticompetitive effects of common ownership or that there is no evidence of it either do not attempt to refute AST’s findings of anticompetitive effects in the U.S. airlines industry or in fact confirm the evidence by AST and even dispel valid concerns about AST’s methodology. Focusing on Kennedy, O’Brien, Song, and Waehrer (KOSW), we note that their panel regressions using market-share-free indices of common ownership concentration confirm the positive correlation between common ownership concentration and price, which AST showed with a measure containing potentially endogenous market shares. We then examine the alternative empirical methods KOSW propose: (i) Their conclusion that estimates from a structural model show no evidence of anticompetitive effects is based on an estimation that discards 90% of the available data and therefore, at best, is only valid for that subsample; (ii) their structural model makes no economic sense because it produces a negative effect of route distance on marginal cost; and (iii) they construct an alternative version of the widely used BlackRock- Barclays Global Investors instrument that is arguably invalid. Even absent these methodological concerns, KOSW’s structural estimates are so noisy that they do not in fact reject the hypothesis that common ownership concentration has a positive effect on prices. A more recent structural paper by Park and Seo has shown these concerns to be well-founded: using a different and larger subsample of AST’s data and more standard estimation methods compared to KOSW, they estimate a positive effect of common ownership on prices, as well as a positive effect of route distance on cost. A lesson for future research—and readers of the literature—is to critically evaluate the conclusions drawn by studies in this field, including those that advertise themselves as providing evidence against the existence of anticompetitive effects of common ownership.